IJRET: International Journal of Research in Engineering and Technology

eISSN: 2319-1163 | pISSN: 2321-7308

Rajeshwari A1, Mani N D2
Assistant Professor, Department of Rural Development, The Gandhigram Rural Institute – Deemed University, Dindigul,
Tamil Nadu, India
Professor, Department of Rural Development, The Gandhigram Rural Institute – Deemed University, Dindigul, Tamil
Nadu, India


Land Surface Temperature (LST) is an important phenomenon in global climate change. As the green house gases in the atmosphere
increases, the LST will also increase. This will result in melting of glaciers and ice sheets and affects the vegetation of that area. Its
impact will be more in the monsoon areas, because the rainfall is unpredictable, failure of monsoon and there will be heavy down
pour of rainfall. LST can be estimated through many algorithms viz., Split-Window (SW), Dual-Angle (DA), Single-Channel (SC),
Sobrino and Mao. With the advent of satellite images and digital image processing software, now it is possible to calculate LST. In
this study, LST for Dindigul District, Tamil Nadu, India, was derived using SW algorithm with the use of Landsat 8 Optical Land
Imager (OLI) of 30 m resolution and Thermal Infrared Sensor (TIR) data of 100 m resolution. SW algorithm needs spectral radiance
and emissivity of two TIR bands as input for deriving LST. The spectral radiance was estimated using TIR bands 10 and 11. Emissivity
was derived with the help of NDVI threshold technique for which OLI bands 2, 3, 4 and 5 were used. The output revealed that LST
was high in the barren regions whereas it was low in the hilly regions because of vegetative cover. As the SW algorithm uses both the
TIR bands (10 and 11) and OLI bands 2, 3, 4 and 5, the LST generated using them were more reliable and accurate.

Keywords: Land surface temperature, Split-window algorithm, OLI, TIR, NDVI.
---------------------------------------------------------------------***--------------------------------------------------------------------1. INTRODUCTION
Land Surface Temperature (LST) can be defined as the
temperature felt when the land surface is touched with the
hands or it is the skin temperature of the ground [1, 2, 3]. LST
is the temperature emitted by the surface and measured in
kelvin. It was greatly affected by the increasing green house
gases in the atmosphere. As it rises, it melts the glaciers and
ices sheets in the polar region. Thus it leads to flood and sea
level rise. Increase in LST also affects the climatic condition
of the monsoon countries leading to unpredictable rainfall.
The vegetation in the entire Earth surface will be affected by
this. Land use/ Land cover (LU/LC) of an area can be used for
estimating the amount of LST. The natural and anthropogenic
activities change the LU/LC of an area. This also influences
LST of that area. As its value changes the local climate of the
area also changes. It is an important phenomenon to be
investigated. Hence, many researchers had calculated LST
using various algorithms and techniques.
Before the invention of Earth Observation Satellites (EOS), it
was hard to estimate the LST of an area. Generally, it was
calculated for a particular set of sample points and interpolated
into isotherms to generalize the point data into area data. Now
with the advent of satellites and high resolution sensors it is
possible to estimate LST spatially. It can be calculated for a
region at a stretch with the use of thermal infrared bands

supplied by satellites. Landsat 8 comes with two different sets
of images from Operation Land Imager (OLI) sensor with nine
bands (band 1 to 9) and Thermal Infrared sensor (TIR) with
two bands (band 10 and 11).

1.1 Literature Reviewed
LST was generated using MODIS (31 and 32 bands) and
ASTER data of same date and time for one of the semi-arid
region in Iran. LST output of MODIS exactly matches with
the ASTER output [4]. Mono-window and single channel
algorithm were used to estimate LST for Alashtar city, Iran.
Emissivity was calculated using NDVI and supervised
classification method of LU/LC. LST was compared with in
situ and the result showed a positive correlation with NDVI
and LU/LC method [5]. NDVI was found for four different
years in Kunsan city Chollabuk_do, Korea. Using NDVI
threshold technique LST was derived. The output of LST was
compared with the NDVI output and a positive correlation was
found in between them. Changes over period was also
assessed and found that LST was increasing [6]. LST and
NDVI were estimated using Landsat 7 image. The Fractional
Vegetation Cover (FVC) for each pixel was calculated and it
was used in LST analysis. A correlation study was conducted
between LST and NDVI data. The output revealed that there
was a strong positive correlation between them and the

Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org


IJRET: International Journal of Research in Engineering and Technology

methodology is more feasible to estimate NDVI and surface
emissivity [7].
Single –Window algorithm for estimating LST was adjusted
for Landsat 8 data for better accuracy. The basic inputs for
SW algorithm were brightness temperature and Land Surface
Emissivity (LSE). The adjusted SW algorithm’s accuracy was
estimated using MODTRAN 4.0 software. The study was
conducted in the northern Negev Desert, Israel [8]. Sobrino
and Mao methods were used individually for retrieving LST
with MODIS data in Hebei and Shanxi, North China Plain.
The maximum, minimum and mean of Sobrino and Mao LST
methods were cross checked with the standard LST values and
found that Sobrino output range greater while Mao method
had less value than standard LST. Hence, a combined method
of Sobrino and Mao was evolved as Sobmao method. It is as
accurate as Sobrino and simple to use [9]. Along-Track
Scanning Radiometer-2 (ATSR-2) data was processed in
MODTRAN 3.5 simulations. Split-Window, dual-angle and
mixed structures algorithms were used to generate LST and
Sea Surface Temperature (SST) in a part of New South Wales.
The dual-angle algorithm has low sensitive to aerosol effects
Thus many researchers had estimated LST using satellite
image. A number of algorithms were developed and adopted
by them to estimate LST. Some of the frequently used
algorithms are Split-Window (SW), Sobrino, Mao, DualAngle, and Sobmao. Most of the studies were done for urban
areas and arid and semi-arid regions and in many of the
studies, single thermal band was used. In the present study
LST was estimated for the entire district using two TIR bands
and four OLI bands.

eISSN: 2319-1163 | pISSN: 2321-7308

estimate brightness temperature and OLI spectral bands 2, 3, 4
and 5 were used to generate NDVI of the study area. The
district was covered in four tiles. Landsat 8 provides metadata
of the bands such as thermal constant, rescaling factor value
etc., which can be used for calculating various algorithms like
Table – 1: Metadata of Satellite Images

No. of


Path/ Row

Date of
24th March

Table – 2: K1 and K2 Values
Thermal Constant

Band 10


Table - 3: Rescaling Factor
Rescaling Factor

Band 10

Band 11

2.1 Flowchart

The major objectives of the study are to find the brightness
temperature using band 10 and band 11 of TIR, calculate the
LSE using NDVI threshold technique and estimate the LST of
Dindigul district using Split-Window (SW) algorithm.

1.2 Study Area
Dindigul district is one of the drought prone districts in Tamil
Nadu State, India. The district lies between 10o0’ 50” N to
10o50’35” N latitude and 77o15’53”E to 78o21’2”E longitude.
The district consists of Palani and Kodaikanal hills in the
south west, a few small hills viz., Sirumalai and Karanthmalai
in the south east. Northern part of the district is undulating
plain and majority of the area was under rainfed cultivation.
Due to frequent failure of monsoon, the district faces severe
drought since 2012. Hence, this district has been selected
purposively for the present study.

1.3 Data and its Source
Landsat 8 is one of the Landsat series of NASA. The data of
Landsat 8 is available in Earth Explorer website at free of cost.
In the present study, the TIR bands 10 and 11 were used to

Fig - 1: Flow Chart

Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org


IJRET: International Journal of Research in Engineering and Technology

eISSN: 2319-1163 | pISSN: 2321-7308

2.2 Process

2.2.2 Top of Atmospheric Spectral Radiance

LST was calculated by applying a structured mathematical
algorithm viz., Split-Window (SW) algorithm. It uses
brightness temperature of two bands of TIR, mean and
difference in land surface emissivity for estimating LST of an
area. The algorithm is

The value of Top of Atmospheric (TOA) spectral radiance
(Lλ) was determined by multiplying multiplicative rescaling
factor (0.000342) of TIR bands with its corresponding TIR
band and adding additive rescaling factor (0.1) with it.
Lλ = ML*Qcal + AL

LST = TB10 + C1 (TB10-TB11) + C2 (TB10-TB11)2 + C0 +
(C3+C4W) (1- ε) + (C5+C6W) ∆ ε


LST - Land Surface Temperature (K)
C0 to C6 - Split-Window Coefficient values (table – 4)
(Skokovic et al, 2014; Sobrino et al, 1996; 2003; Shaouhua
Zhao et al, 2009)
TB10 and TB11 – brightness temperature of band 10 and band
11 (K)
ε – mean LSE of TIR bands
W – Atmospheric water vapour content
∆ ε – Difference in LSE
Table – 4: SW Coefficient Values


2.2.3 Land Surface Emissivity
To find LST it is necessary to calculate the LSE of the region.
LSE was estimated using NDVI threshold method.
LSE = εs (1-FVC) + εv * FVC


εs and εv - soil and vegetative emissivity values of the
corresponding bands.


Table – 5: Emissivity Values [11]

Brightness temperature (TB) is the microwave radiation
radiance traveling upward from the top of Earth's
atmosphere. The calibration process has been done for
converting thermal DN values of thermal bands of TIR to TB.
For finding TB of an area the Top of Atmospheric (TOA)
spectral radiance of (Lλ) was needed. TB for both the TIRs
bands was calculated by adopting the following formula,

Lλ - Top of Atmospheric Radiance in watts/ (m2*srad*µm)
ML - Band specific multiplicative rescaling
factor (radiance_mult_band_10/11)
Qcal - band 10/ 11 image.
AL - Band specific additive rescaling factor


2.2.1 Brightness Temperature

TB =



K1 and K2- thermal conversion constant and it varies for both
TIR bands (table 2)
Lλ – Top of Atmospheric spectral radiance.

Band 10

Band 11

FVC - Fractional Vegetation Cover was estimated for a pixel.
FVC for an image was calculated by



NDVIs – NDVI reclassified for soil
NDVIv – NDVI reclassified for vegetation

2.2.4 NDVI Threshold
OLI bands 2, 3, 4 and 5 were layer stacked and NDVI was
calculated using ERDAS Imagine software. The output value
of NDVI ranged between -1 and 0.59. To get NDVIs and
NDVIv, the NDVI image was reclassified into soil and
vegetation; the classified data were used to find out FVC.
After generating LSE for both the bands of TIR, the mean and
difference LSE was found as,
ε = (ε10- ε11)/2


Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org


IJRET: International Journal of Research in Engineering and Technology
∆ε = ε10- ε11


ε – Mean LSE
∆ε – LSE difference
ε10 and ε11 - LSE of band 10 and 11.
Finally, the LST in kelvin was determined using SW


eISSN: 2319-1163 | pISSN: 2321-7308

Map - 3 has been derived using LSE, brightness temperature
and emissivity difference between LSE of band 10 and 11 of
TIR. LST output portrayed that it varied from less than 305
kelvin to more than 324 kelvin. The highest LST of more than
324 kelvin was traced in the northern plains of the study area,
where barren lands and wastelands were mostly found. The
313 to 318 kelvin LST was traced in the foothill regions and
area under cultivable land in the plains. The lowest category of
less than 305 kelvin was seen in the highly elevated regions
with dense vegetation. Similarly the regions of moderate
vegetation had 305 to 312 kelvin.

NDVI map revealed that the NDVI value ranged between -1 to
0.59. South western part of Dindigul district had highest
NDVI value whereas area under water body had negative
value (Map - 1). The NDVI value of area under vegetation
was more than 0.17 and for built-up and barren land it was 0
to 0.17.

Map – 3

Map - 1
LSE was created using NDVI threshold technique (Map - 2).
The LSE of Dindigul district ranged between 0.97 and 0.988.
Highly elevated regions in the district had more vegetative
cover, hence LSE was high in these regions. High LSE was
found in southern, south eastern and south western parts of the
district, whereas low LSE was noticed in northern and central
parts of the study area.

LST of an area was determined based on its brightness
temperature and LSE using SW algorithm. In this study, OLI
and TIR bands of Landsat 8 had been used. The study clearly
revealed that as the district had more vegetative cover in hilly
regions the LST in southern part was low and the northern
plains with barren lands, uncultivable land and urban areas
experienced high LST. Dindigul district being a drought prone
district the area under vegetation was less and it is restricted to
hilly areas in the south western and south eastern parts. Thus,
LST can be calculated using SW algorithm on Landsat 8 with
multiband OLI and TIR images.





Map - 2

Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org


IJRET: International Journal of Research in Engineering and Technology









Akhoondzadeh.M and Saradjian.M.R, “Comparison of
Land Surface Temperature mapping using MODIS and
ASTER Images in Semi-Arid Area”, Commission VIII,
Alipour.T, Sarajian.M.R and Esmaeily.A, “Land
Surface Temperature Estimation from Thermal Band of
Landsat Sensor, Case Study: Alashtar City”, The
International Achieves of the Photogrammetry, Remote
Sensing and Spatial Information Sciences, Vol.384/C7.
Hyung Moo Kim, Beob Kyun Kim and Kang Soo You,
“A Static Correlation Analysis Algorithm between
Land Surface Temperature and Vegetative Index”,
International Journal of Information Processing
Systems, Vol.1, No.1, pp 102-106, 2005.
Jovet Mallick, Yogesh Kant and Bharath.B.D,
“Estimation of Land Surface Temperature Over Delhi
using Landsat-7 ETM+”, Journal of Indian Geophysics
Union, Vol.12, No.3, pp 131-140, July 2008.
Offer Rozenstein, Zhihao Qin, Yevgency Derimian and
Arnon Kornieli, “Derivation of
Land Surface
Temperature for Landsat-8 TIRs using a Split-Window
Algrithm”, Sensors, vol.14, pp 5768-5780, 2014.
Shaohua Zhao, Qiming Qin, Yonghui Yang, Yujiu
Xiong and Guoyu Qiu, “Comparison og two SplitWindow Mehtods for Retrieving Land Surface
Temperature from MODIS Data”, Journal of Earth
Syst. Science, Vol.118, No.4, pp 345-353, August
Sobrino.J.A, Reillo.S, Cueca.J and Prata.A.J,
“Algorithms for Estimating Surface Temperature from
Skokovic.D, Sobrino.J.A, Jimenez-Munoz.J.C, Soria.G,
Julien.Y, Mattar.C and Jordi Cristobal, “Calibration
and Validation of Land Surface Temperature for
Landsat 8 – TIRS Sensor”, Land product Validation
and Evolution, ESA/ESRIN Frascati (Italy), pp 6-9,
January 28-30, 2014.

eISSN: 2319-1163 | pISSN: 2321-7308

Mani.N.D took M.Sc.Geography in 1979 and
Ph.d in Perceptional Cartography in 1994. He
is a Professor in the Department of Rural
Development, the Gandhigram Rural Institute
– Deemed University since 1983. He has
successfully completed 32 research projects in
the areas of digital cartography, application of Geographical
Information System and remote sensing in natural resource

Rajeshwari.A obtained M.Sc degree in
Geoinformatics in 2006 and Ph.D in GIS and
Satellite Imageries Aided Impact Assessment
of Watershed Development Projects in
Dindigul District in 2010. She is currently
working as Assistant Professor, since 2010, in Geoinformatics,
in the Department of Rural Development, the Gandhigram
Rural Institute (Deemed University). Her area of specialisation
are watershed, application of Geographical Information
System and Remote Sensing.

Volume: 03 Issue: 05 | May-2014, Available @ http://www.ijret.org